Risk of cardiovascular disease in relation to the use of combined postmenopausal hormone therapy: detection bias and resolution of discrepant findings in two Women's Health Initiative studies

Abstract

Context In a recent analysis of data in the Women's Health Initiative (WHI) clinical trial and observational studies, the overall risks of coronary heart disease, stroke and venous thromboembolism, in postmenopausal women who used combined therapy with conjugated estrogens and medroxyprogesterone acetate, were lower in the observational data. However, the risks became more similar within similar duration categories of follow-up. In both studies, there was initial elevation, followed by a duration-dependent reduction, in the risks. The investigators suggest that the discrepancies in the overall risks may be due to weighting by short-duration exposure in the clinical trial, and long-duration exposure in the observational study.

Critique In the clinical trial, 44% of the hormone therapy recipients were unblinded, mainly because of persistent vaginal bleeding, and awareness of exposure status was therefore common; in the observational study, all exposed women were aware. When follow-up in the WHI studies commenced, extensive publicity was given to evidence to suggest that, among women who had already sustained a myocardial infarction, combined hormone therapy may not decrease the risk of a second myocardial infarct, as hypothesized, but instead increase it. From the commencement of follow-up onward, detection bias was therefore possible in both WHI studies. That possibility was further reinforced after about 3 years of follow-up, and again a year later, when the clinical trial participants were explicitly informed that supplemental hormones were associated with increased cardiovascular risks. That information was also given extensive publicity.

Conclusions Detection bias could not be ruled out in either WHI study, and there may have been systematic overestimation of the risks of cardiovascular outcomes associated with the use of combined hormone therapy. To a limited extent, an analysis of the clinical trial data according to blinding status might be informative in determining whether detection bias could have given rise to overestimation of the overall and duration-specific risks in both WHI studies. If numbers permit, comparisons within strata of severity might also be informative. If the risks were overestimated in the WHI studies, a duration-related protective effect of combined hormone therapy on the risk of coronary heart disease and stroke may have gone undetected.

Key words:

• CARDIOVASCULAR DISEASE
• POSTMENOPAUSAL HORMONE THERAPY
• DETECTION BIAS
• WOMEN'S HEALTH INITIATIVE